作者selient (假安靜)
站內Statistics
標題[程式] SAS 存活分析 cox proportional hazards
時間Tue Apr 13 19:31:58 2010
------------------------------------------------------------------------
[軟體程式類別]: sas
[程式問題]: cox proportional hazards 資料處理
[軟體熟悉度]:
高(1年以上)
[問題敘述]:
目的是根據不同的treatment( trt ) 來做
PHREG
然後在同一張圖上畫出兩條 LOG-LOG survival curve
請問有沒有辦法不用
WHERE 這個指令
因為這樣不同的trt必須分開做
然後在用
SET把資料彙整
請問有別的指令可以用嗎？
[程式範例]:
DATA r;
INPUT group weeks status wbc sex@@;
trt=group-1;
IF wbc<2.3 THEN lmh=1;
ELSE IF wbc>3 THEN lmh=3;
ELSE lmh=2;
CARDS;
1 6 1 2.31 0 1 6 1 4.06 1 1 6 1 3.28 0
1 7 1 4.43 0 1 10 1 2.96 0 1 13 1 2.88 0
1 16 1 3.60 1 1 22 1 2.32 1 1 23 1 2.57 1
1 6 0 3.20 0 1 9 0 2.80 0 1 10 0 2.70 0
1 11 0 2.60 0 1 17 0 2.16 0 1 19 0 2.05 0
1 20 0 2.01 1 1 25 0 1.78 1 1 32 0 2.20 1
1 32 0 2.53 1 1 34 0 1.47 1 1 35 0 1.45 1
2 1 1 2.80 1 2 1 1 5.00 1 2 2 1 4.91 1
2 2 1 4.48 1 2 3 1 4.01 1 2 4 1 4.36 1
2 4 1 2.42 1 2 5 1 3.49 1 2 5 1 3.97 0
2 8 1 3.52 0 2 8 1 3.05 0 2 8 1 2.32 0
2 8 1 3.26 1 2 11 1 3.49 0 2 11 1 2.12 0
2 12 1 1.50 0 2 12 1 3.06 0 2 15 1 2.30 0
2 17 1 2.95 0 2 22 1 2.73 0 2 23 1 1.97 1
;
DATA inrisk1;
wbc=2.93;
PROC PHREG DATA=remission NOPRINT;
MODEL weeks*status(0)=wbc;
WHERE trt=0;
BASELINE COVARIATES=inrisk1 OUT=out10 LOGLOGS=lls/NOMEAN;
PROC PHREG DATA=remission NOPRINT;
MODEL weeks*status(0)=wbc;
WHERE trt=1;
BASELINE COVARIATES=inrisk1 OUT=out11 LOGLOGS=lls/NOMEAN;
DATA out1;
SET out10 (IN=in0) out11 (IN=in1);
IF in0 THEN trt=0;
ELSE IF in1 THEN trt=1;
lls=-lls;
PROC PRINT;
-----------------------------------------------------------------------------
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 140.119.137.219
→ selient:噢 畫出LLS的指令忘記打，不過重點不是那個請見諒 04/13 19:34
推 kenshin528:BY ? 04/13 19:50
→ selient:用過 by 不行耶 print不出來東西 04/13 20:49
推 wlsherica:STRATA可嗎???? 04/13 21:11
→ bmka:I think what you want is fitting Kaplan-Meier curves for 04/13 23:19
→ bmka:different treatment groups. 04/13 23:20
→ selient:not KM , I want PH 04/13 23:58
→ bmka:It's not very meaningful because you already impose a 04/14 00:45
→ bmka:proportional hazards assumption. 04/14 00:45
→ selient:under that assumption I draw two curve to check if 04/14 01:05
→ selient:it fit the assumption. 04/14 01:06
→ bmka:Then what you want is KM curves. 04/14 01:33
→ bmka:or other model diagnosis plots. Keep in mind that you 04/14 01:35
→ bmka:are dealing with censored data. 04/14 01:35